A Similarity Measure for Automatic Audio Classi cation

This paper presents recent results using statistics generated by a MMI-supervised vector quantizer as a measure of audio similarity. Such a measure has proved successful for talker identi cation, and the extension from speech to general audio, such as music, is straightforward. A classi er that distinguishes speech from music and non-vocal sounds is presented, as well as experimental results showing how perfect classi cation accuracy may be achieved on a small corpus using substantially less than two seconds per test audio le. The techniques a presented here may be extended to other applications and domains, such as audio retrieval-by-similarity, musical genre classi cation, and automatic segmentation of continuous audio.